At its Cloud Next 2025 event, Google unveiled Ironwood — its 7th-generation TPU — and a new lineup of AI models aimed at shaping what it calls the “age of inference,” where AI doesn’t just respond, it thinks ahead.
AI-generated music, realistic voices, and smarter transcription
Google’s Lyria model, now in enterprise preview, turns text prompts into high-fidelity music across genres — custom soundtracks in minutes, without licensing headaches. “Lyria eliminates these hurdles, allowing you to generate custom music tracks in minutes, directly aligning with your content's mood, pacing, and narrative,” said Google. Lyria will be available for Google Cloud enterprise users.
Meanwhile, Chirp 3, the upgraded audio model, adds HD voice synthesis in 35+ languages and the ability to clone a voice from just 10 seconds of audio.
A new transcription feature separates voices in multi-speaker audio — ideal for call centers and podcasts. “You can also weave AI-powered narration into your existing recordings, and add a speech transcription capability that can distinguish between speakers,” said Google.
Google also said that to ensure responsible use, Instant Custom Voice includes built-in safety features, and its onboarding process involves rigorous diligence to verify proper voice usage permissions.
Finally, to keep things ethical, DeepMind’s SynthID now watermarks every image, video, and audio frame generated by Google’s creative models.
New Gemini models for real-world AI
Google also introduced Gemini 2.5 Flash, a new “workhorse” model focused on low-latency, cost-sensitive applications like customer support. It can auto-adjust its “thinking time” based on task complexity — or let users manually fine-tune the trade-off between speed, accuracy, and price, as per Google.
Discover the latest Business News, Sensex, and Nifty updates. Obtain Personal Finance insights, tax queries, and expert opinions on Moneycontrol or download the Moneycontrol App to stay updated!
